DEPARTMENT OF THE AIR FORCE AIR UNIVERSITY AIR FORCE INSTITUTE OF TECHNOLOGY Wright-Patterson Air Force Base, Ohio APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED AN INVESTIGATION RELATING LONGITUDINAL PILOT-INDUCED OSCILLATION TENDENCY RATING TO DESCRIBING FUNCTION PREDICTIONS FOR RATE-LIMITED ACTUATORS THESIS Joel B. Witte, Major, USAF AFIT/GAE/ENY/04-M16
154
Embed
AN INVESTIGATION RELATING LONGITUDINAL PILOT-INDUCED ... · 4-1. MAX GAP (LAMARS) Nichols Charts for Cases B, N, W and Y.....4-3 4-2. MAX GAP (LAMARS) Phase 2 Sum-of-sines Task Data.....4-6
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
DEPARTMENT OF THE AIR FORCE AIR UNIVERSITY
AIR FORCE INSTITUTE OF TECHNOLOGY
Wright-Patterson Air Force Base, Ohio
APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED
AN INVESTIGATION RELATING LONGITUDINAL PILOT-INDUCED
OSCILLATION TENDENCY RATING TO DESCRIBING FUNCTION
PREDICTIONS FOR RATE-LIMITED ACTUATORS
THESIS
Joel B. Witte, Major, USAF
AFIT/GAE/ENY/04-M16
The views expressed in this thesis are those of the author and do not reflect the official policy or position of the United States Air Force, Department of Defense, or the United States Government
AFIT/GAE/ENY/04-M16
AN INVESTIGATION RELATING LONGITUDINAL PILOT-INDUCED
OSCILLATION TENDENCY RATING TO DESCRIBING FUNCTION
PREDICTIONS FOR RATE-LIMITED ACTUATORS
THESIS
Presented to the Faculty
Department of Aeronautics and Astronautics
Graduate School of Engineering and Management
Air Force Institute of Technology
Air University
Air Education and Training Command
In Partial Fulfillment of the Requirements for the
Degree of Master of Science in Aeronautical Engineering
Joel B. Witte, BS
Major, USAF
March 2004
APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED.
AFIT/GAE/ENY/04-M16
AN INVESTIGATION RELATING LONGITUDINAL PILOT-INDUCED
OSCILLATION TENDENCY RATING TO DESCRIBING FUNCTION
PREDICTIONS FOR RATE-LIMITED ACTUATORS
Joel B. Witte, BS Major, USAF
Approved: __________________________________ ______________ Bradley S. Liebst (Chairman) date __________________________________ ______________ Richard G. Cobb (Member) date __________________________________ ______________ Russell G. Adelgren (Member) date
iv
AFIT/GAE/ENY/04-M16
Abstract
The purpose of this study was to investigate pilot-induced oscillations (PIO) and
determine a method by which a PIO tendency rating could be predicted. In particular,
longitudinal PIO in the presence of rate-limited actuators were singled out for
examination. Sinusoidal input/triangular output describing function techniques using
Nichols charts were used. A new criterion dubbed Gap Criterion was calculated for PIO
sensitivity. This criterion consists of the product of additional pilot gain and the
normalized maximum amplitude of the commanded actuator necessary to cause PIO.
These results were paired with simulator and flight test PIO tendency rating data. The
PIO rating scale used was the PIO tendency classification of MIL-HDBK-1797. This
concept was applied to two historical test databases, HAVE PREVENT and HAVE
OLOP. Additional PIO data was gathered in the Large Amplitude Multimode Aerospace
Simulator (LAMARS) at the Air Force Research Laboratory (AFRL), Wright-Patterson
AFB, Ohio and the USAF NF-16D Variable In-flight Stability Test Aircraft (VISTA) at
Edwards AFB, California. Correlation between PIO tendency rating and Gap Criterion
was determined for each dataset. Most datasets exceeded a confidence level of 95% that
a correlation existed. Follow-on analysis for better curve fitting was accomplished; a
logarithmic fit was judged best. Datasets were combined with success to demonstrate the
universality of the Gap Criterion for correlating PIO tendency ratings for longitudinal
PIO involving rate-limited actuators.
v
Acknowledgements
The Air Force Institute of Technology and the United States Air Force Test Pilot
School cosponsored this study. I would like to thank the following people for their
efforts and contributions: My thesis advisor, Dr Brad Liebst, who proposed this study and
offered key insights; Mr. Andy Markofski and Mr. Mike Steen of General Dynamics
Advanced Information Systems who programmed the aircraft dynamics for the VISTA
NF-16D; the VISTA maintenance team who worked miracles to return the aircraft to
flying status; Major Russ Adelgren and Mrs. Cynthia Roell, who made sure the flight test
portion succeeded and Mr. Curt Clark and Mr. Jeff Slutz who worked diligently to ready
the LAMARS simulator for the simulator test portion.
I’d especially like to thank the members of the MAX GAP test team who made
this project succeed: Captain Erik Monsen, Royal Norwegian Air Force; Captain Thomas
PIO Defined. ............................................................................................................ 1-2 PIO History. ............................................................................................................. 1-4 PIO Categories. ........................................................................................................ 1-6 PIO Category II........................................................................................................ 1-7 PIO Scales................................................................................................................ 1-8 PIO Prediction Methods......................................................................................... 1-10 Gap Criterion. ........................................................................................................ 1-11
II. Theory ...................................................................................................................... 2-1
Describing Function Development .............................................................................. 2-1 Saturation Nonlinearity Describing Function.............................................................. 2-3 Closed Loop Describing Function Approximation...................................................... 2-4 Sinusoidal Input/Triangle Output Describing Function Approximation ..................... 2-7 Applying Describing Function Results to Predict PIO.............................................. 2-11 Pilot Model ................................................................................................................ 2-13 Gap Criterion............................................................................................................. 2-16
Gap Criterion Formulation..................................................................................... 2-17 Type I. .................................................................................................................... 2-19 Type II.................................................................................................................... 2-20 Type III. ................................................................................................................. 2-21 Type IV. ................................................................................................................. 2-21
vii
Page
Example of Gap Criterion Application ..................................................................... 2-21 III. Analysis of Selected Historical Data ...................................................................... 3-1
HAVE PREVENT Analysis ........................................................................................ 3-1 Gap Criterion Calculation for HAVE PREVENT Datasets..................................... 3-4 HAVE PREVENT Gap Criterion Summary............................................................ 3-7 HAVE PREVENT Gap Criterion Correlation. ........................................................ 3-8 HAVE PREVENT Best Curve Fit. ........................................................................ 3-11
HAVE OLOP Analysis.............................................................................................. 3-14 Gap Criterion Calculation for HAVE OLOP Datasets. ......................................... 3-15 HAVE OLOP Gap Criterion Summary. ................................................................ 3-19 HAVE OLOP Gap Criterion Correlation............................................................... 3-19 HAVE OLOP Best Curve Fit................................................................................. 3-21
HAVE PREVENT and HAVE OLOP Summary ...................................................... 3-24 IV Analysis of Project MAX GAP (LAMARS) Data ................................................... 4-1
MAX GAP (LAMARS) Analysis................................................................................ 4-1 Gap Criterion Calculation for MAX GAP (LAMARS) Datasets. ........................... 4-2 MAX GAP (LAMARS) Gap Criterion Summary. .................................................. 4-5 MAX GAP (LAMARS) Gap Criterion Correlation................................................. 4-5 MAX GAP (LAMARS) Best Curve Fit................................................................... 4-8
MAX GAP (LAMARS) Summary ............................................................................ 4-11 V. Analysis of Project MAX GAP (VISTA) Data........................................................ 5-1
MAX GAP (VISTA) Analysis..................................................................................... 5-1 Gap Criterion Calculation for MAX GAP (VISTA) Datasets. ................................ 5-2 MAX GAP (VISTA) Gap Criterion Summary. ....................................................... 5-6 MAX GAP (VISTA) Gap Criterion Correlation. .................................................... 5-6 MAX GAP (VISTA) Best Curve Fit. ...................................................................... 5-8
MAX GAP (VISTA) Summary ................................................................................. 5-10 VI. Analysis of Combined Gap Criterion Data ............................................................ 6-1
LAMARS Combined Datasets .................................................................................... 6-1 VISTA Combined Datasets ......................................................................................... 6-3 LAMARS and VISTA Combined Dataset Correlation ............................................... 6-4 Combined Dataset Analysis and Observations............................................................ 6-5
VII. Conclusions ........................................................................................................... 7-1
viii
Page
Appendix A. Matlab/SimulinkTM Code ......................................................................... A-1
Appendix B. State-Space Matrices for Project MAX GAP........................................... B-1
Appendix C. Correlation Computation .......................................................................... C-1
Appendix D. MAX GAP Histograms ............................................................................ D-1
Figure Page 1-1. YF-22A Accident Sequence (Hodgkinson, 1999:128)............................................ 1-4 1-2. Actuator Saturation Example Using an Input = A sin(ωi t) ..................................... 1-8 1-3. SimulinkTM Actuator Model .................................................................................... 1-8 1-4. PIO Tendency Classification (MIL-HDBK-1797, 1997:152)................................. 1-9 2-1. Example of a Nonlinear System .............................................................................. 2-1 2-2. Saturation Nonlinearity and the Corresponding Input-Output Relationship (Slotine
and Li, 1995:173)..................................................................................................... 2-3 2-3. Actuator Model Development (Klyde and others, 1995:22) ................................... 2-5 2-4. Closed Loop Actuator Transfer Function Diagram................................................. 2-6 2-5. Rate-Limiting Input and Output (Klyde and others, 1995:42-46)........................... 2-8 2-6. Describing Function Phase Angle Comparison..................................................... 2-10 2-7. Pitch Tracking Closed Loop System ..................................................................... 2-11 2-8. Simplified Pitch Tracking Closed Loop System ................................................... 2-11 2-9. Neal-Smith Pilot Model Constraints...................................................................... 2-15 2-10. Pitch Tracking Closed Loop System for Gap Criterion...................................... 2-16 2-11. Four Resulting Gap Criterion Types................................................................... 2-18 2-12. Case I Effective Pilot Gain Increase.................................................................... 2-19 2-13. Case II Effective Pilot Gain Decrease ................................................................. 2-20 2-14. Nichols Chart of the Example Problem............................................................... 2-22 3-1. Phase 2 Sum-of-sines Pitch Tracking Task ............................................................. 3-2 3-2. Phase 3 Discrete HUD Pitch Tracking Task ........................................................... 3-2
x
Figure .............................................................................................................................Page 3-3. Phase 3 Target Tracking Task ................................................................................. 3-3 3-4. Longitudinal State Space Diagram (Matlab/SimulinkTM, 2001) ............................. 3-4 3-5. HAVE PREVENT Nichols Charts for Cases A, B and C ....................................... 3-6 3-6. HAVE PREVENT Phase 2 Sum-of-sines Task LAMARS Data ............................ 3-9 3-7. HAVE PREVENT Phase 3 Discrete HUD Pitch Tracking Task LAMARS Data .. 3-9 3-8. HAVE PREVENT Phase 3 Target Tracking Task LAMARS Data...................... 3-10 3-9. HAVE PREVENT Phase 2 Sum-of-sines Task LAMARS Data with Logarithmic
Curve Fit ................................................................................................................ 3-12 3-10. HAVE PREVENT Phase 3 Discrete HUD Pitch Tracking Task LAMARS Data
with Logarithmic Curve Fit ................................................................................... 3-13 3-11. HAVE PREVENT Phase 3 Target Tracking Task LAMARS Data with
Logarithmic Curve Fit............................................................................................ 3-13 3-12. HAVE OLOP Nichols Charts for Cases A, B and C........................................... 3-17 3-13. HAVE OLOP Phase 2 Sum-of-sines Task VISTA Data..................................... 3-20 3-14. HAVE OLOP Phase 3 Discrete HUD Pitch Tracking Task VISTA Data........... 3-20 3-15. HAVE OLOP Phase 2 Sum-of-sines Task VISTA Data with Logarithmic Fit... 3-23 3-16. HAVE OLOP Phase 3 Discrete HUD Pitch Tracking Task VISTA Data with
Logarithmic Fit ...................................................................................................... 3-23 4-1. MAX GAP (LAMARS) Nichols Charts for Cases B, N, W and Y......................... 4-3 4-2. MAX GAP (LAMARS) Phase 2 Sum-of-sines Task Data...................................... 4-6 4-3. MAX GAP (LAMARS) Phase 3 Discrete HUD Pitch Tracking Task Data ........... 4-6 4-4. MAX GAP (LAMARS) Phase 3 Target Tracking Task Data ................................. 4-7 4-5. MAX GAP (LAMARS) Phase 2 Sum-of-sines Task Data with Logarithmic Curve
Fit ............................................................................................................................. 4-9
xi
Figure .............................................................................................................................Page 4-6. MAX GAP (LAMARS ) Phase 3 Discrete HUD Pitch Tracking Task Data with
Logarithmic Fit ...................................................................................................... 4-10 4-7. MAX GAP (LAMARS) Phase 3 Target Tracking Task Data with Logarithmic
Fit ........................................................................................................................... 4-10 5-1. MAX GAP (VISTA) Nichols Charts for Cases B, N, W and Y.............................. 5-4 5-2. MAX GAP (VISTA) Phase 2 Sum-of-sines Task Data .......................................... 5-7 5-3. MAX GAP (VISTA) Phase 3 Discrete HUD Pitch Tracking Task Data ................ 5-7 5-4. MAX GAP (VISTA) Phase 2 Sum-of-sines Task Data with Logarithmic Curve
Fit ............................................................................................................................5--9 5-5. MAX GAP (VISTA) Phase 3 Discrete HUD Pitch Tracking Task Data with
Logarithmic Curve Fit............................................................................................ 5-10 6-1. LAMARS Combined Phase2 Sum-of-sines Dataset ............................................... 6-1 6-2. LAMARS Combined Phase 3 Discrete (HUD) Pitch-Tracking Dataset ................. 6-2 6-3. LAMARS Combined Phase3 Target Tracking Dataset........................................... 6-2 6-4. VISTA Combined Phase 2 Sum-of-sines Dataset ................................................... 6-3 6-5. VISTA Combined Phase 3 Discrete (HUD) Pitch-Tracking Dataset...................... 6-3 6-6. LAMARS and VISTA Combined Phase 2 Sum-of-sines Data ............................... 6-4 6-7. LAMARS and VISTA Combined Phase 3 Discrete Pitch-Tracking Data .............. 6-4 A-1. Augmented Dynamics SimulinkTM Model ............................................................ A-4 D-1. LAMARS Phase 2 Sum-of-sines Data, Case B, 15 deg/sec .................................. D-1 D-2. LAMARS Phase 2 Sum-of-sines Data, Case B, 30 deg/sec .................................. D-2 D-3. LAMARS Phase 2 Sum-of-sines Data, Case B, 60 deg/sec .................................. D-2 D-4. LAMARS Phase 2 Sum-of-sines Data, Case N, 15 deg/sec.................................. D-3 D-5. LAMARS Phase 2 Sum-of-sines Data, Case N, 30 deg/sec.................................. D-3
xii
Figure .............................................................................................................................Page D-6. LAMARS Phase 2 Sum-of-sines Data, Case N, 60 deg/sec.................................. D-4 D-7. LAMARS Phase 2 Sum-of-sines Data, Case W, 15 deg/sec................................. D-4 D-8. LAMARS Phase 2 Sum-of-sines Data, Case W, 30 deg/sec................................. D-5 D-9. LAMARS Phase 2 Sum-of-sines Data, Case W, 60 deg/sec................................. D-5 D-10. LAMARS Phase 2 Sum-of-sines Data, Case Y, 15 deg/sec................................ D-6 D-11. LAMARS Phase 2 Sum-of-sines Data, Case Y, 30 deg/sec................................ D-6 D-12. LAMARS Phase 2 Sum-of-sines Data, Case Y, 60 deg/sec................................ D-7 D-13. LAMARS Phase 3 Discrete HUD Pitch Tracking Data, Case B, 15 deg/sec...... D-8 D-14. LAMARS Phase 3 Discrete HUD Pitch Tracking Data, Case B, 30 deg/sec...... D-8 D-15. LAMARS Phase 3 Discrete HUD Pitch Tracking Data, Case B, 60 deg/sec...... D-9 D-16. LAMARS Phase 3 Discrete HUD Pitch Tracking Data, Case N, 15 deg/sec ..... D-9 D-17. LAMARS Phase 3 Discrete HUD Pitch Tracking Data, Case N, 30 deg/sec ... D-10 D-18. LAMARS Phase 3 Discrete HUD Pitch Tracking Data, Case N, 60 deg/sec ... D-10 D-19. LAMARS Phase 3 Discrete HUD Pitch Tracking Data, Case W, 15 deg/sec... D-11 D-20. LAMARS Phase 3 Discrete HUD Pitch Tracking Data, Case W, 30 deg/sec... D-11 D-21. LAMARS Phase 3 Discrete HUD Pitch Tracking Data, Case W, 60 deg/sec... D-12 D-22. LAMARS Phase 3 Discrete HUD Pitch Tracking Data, Case Y, 15 deg/sec ... D-12 D-23. LAMARS Phase 3 Discrete HUD Pitch Tracking Data, Case Y, 30 deg/sec ... D-13 D-24. LAMARS Phase 3 Discrete HUD Pitch Tracking Data, Case Y, 60 deg/sec ... D-13 D-25. LAMARS Phase 3 Target Tracking Data, Case B, 15 deg/sec ......................... D-14 D-26. LAMARS Phase 3 Target Tracking Data, Case B, 30 deg/sec ......................... D-14 D-27. LAMARS Phase 3 Target Tracking Data, Case B, 60 deg/sec ......................... D-15
xiii
Figure .............................................................................................................................Page D-28. LAMARS Phase 3 Target Tracking Data, Case N, 15 deg/sec ......................... D-15 D-29. LAMARS Phase 3 Target Tracking Data, Case N, 30 deg/sec ......................... D-16 D-30. LAMARS Phase 3 Target Tracking Data, Case N, 60 deg/sec ......................... D-16 D-31. LAMARS Phase 3 Target Tracking Data, Case W, 15 deg/sec ........................ D-17 D-32. LAMARS Phase 3 Target Tracking Data, Case W, 30 deg/sec ........................ D-17 D-33. LAMARS Phase 3 Target Tracking Data, Case W, 60 deg/sec ........................ D-18 D-34. LAMARS Phase 3 Target Tracking Data, Case Y, 15 deg/sec ......................... D-18 D-35. LAMARS Phase 3 Target Tracking Data, Case Y, 30 deg/sec ......................... D-19 D-36. LAMARS Phase 3 Target Tracking Data, Case Y, 60 deg/sec ......................... D-19 D-37. VISTA Phase 2 Sum-of-sines Data, Case B, 15 deg/sec................................... D-20 D-38. VISTA Phase 2 Sum-of-sines Data, Case B, 30 deg/sec................................... D-20 D-39. VISTA Phase 2 Sum-of-sines Data, Case B, 60 deg/sec................................... D-21 D-40. VISTA Phase 2 Sum-of-sines Data, Case N, 15 deg/sec................................... D-21 D-41. VISTA Phase 2 Sum-of-sines Data, Case N, 30 deg/sec................................... D-22 D-42. VISTA Phase 2 Sum-of-sines Data, Case N, 60 deg/sec................................... D-22 D-43. VISTA Phase 2 Sum-of-sines Data, Case W, 15 deg/sec.................................. D-23 D-44. VISTA Phase 2 Sum-of-sines Data, Case W, 30 deg/sec.................................. D-23 D-45. VISTA Phase 2 Sum-of-sines Data, Case W, 60 deg/sec.................................. D-24 D-46. VISTA Phase 2 Sum-of-sines Data, Case Y, 15 deg/sec................................... D-24 D-47. VISTA Phase 2 Sum-of-sines Data, Case Y, 30 deg/sec................................... D-25 D-48. VISTA Phase 2 Sum-of-sines Test Data, Case Y, 60 deg/sec........................... D-25 D-49. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case B, 15 deg/sec......... D-26
xiv
Figure .............................................................................................................................Page D-50. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case B, 30 deg/sec......... D-26 D-51. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case B, 30 deg/sec......... D-27 D-52. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case N, 15 deg/sec ........ D-27 D-53. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case N, 30 deg/sec ........ D-28 D-54. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case N, 60 deg/sec ........ D-28 D-55. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case W, 15 deg/sec ....... D-29 D-56. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case W, 30 deg/sec ....... D-29 D-57. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case W, 60 deg/sec ....... D-30 D-58. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case Y, 15 deg/sec ........ D-30 D-59. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case Y, 30 deg/sec ........ D-31 D-60. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case Y, 60 deg/sec ........ D-31
xv
List of Tables
Table Page 1-1. Summary of Famous Longitudinal PIO Events (McRuer, 1995:9)......................... 1-4 1-2. PIO Tendency Rating Scale (MIL-HDBK-1797, 1997:153) ................................ 1-10 2-1. Neal-Smith Pilot Models ....................................................................................... 2-14 3-1. HAVE PREVENT Case Characteristics (Hanley, 2003) ........................................ 3-3 3-2. HAVE PREVENT Pitch-to-Actuator Transfer Functions ( cG ) .............................. 3-4 3-3. HAVE PREVENT Neal-Smith Pilot Models .......................................................... 3-5 3-4. HAVE PREVENT Components for Cases A and B................................................ 3-6 3-5. Gap Criteria Values for HAVE PREVENT Cases A and B.................................... 3-7 3-6. Gap Criteria for HAVE PREVENT Case C ........................................................... 3-7 3-7. HAVE PREVENT Gap Criteria Summary............................................................. 3-8 3-8. HAVE PREVENT Correlation Confidence Levels............................................... 3-11 3-9. HAVE PREVENT Curve Fit Correlation Values.................................................. 3-11 3-10. HAVE OLOP Case Characteristics (Gilbreath, 2001) ........................................ 3-15 3-11. HAVE OLOP Pitch-to-Actuator Transfer Functions ( cG ).................................. 3-15 3-12. HAVE OLOP Neal-Smith Pilot Models.............................................................. 3-16 3-13. HAVE OLOP Components for Cases A and C ................................................... 3-17 3-14. Gap Criteria Values for HAVE OLOP Case A and C......................................... 3-18 3-15. HAVE OLOP Gap Criteria Summary ................................................................ 3-19 3-16. HAVE OLOP Correlation Confidence Levels .................................................... 3-21 3-17. HAVE OLOP Curve Fit Correlation Values ....................................................... 3-22 4-1. MAX GAP (LAMARS) Case Characteristics (Witte and others, 2003)................. 4-1
xvi
Table Page 4-2. MAX GAP (LAMARS) Pitch-to-Actuator Transfer Functions ( cG )...................... 4-2 4-3. MAX GAP (LAMARS) Neal-Smith Pilot Models.................................................. 4-3 4-4. MAX GAP (LAMARS) Components for Cases B, N, W and Y ............................ 4-4 4-5. Gap Criteria Values for MAX GAP (LAMARS) Cases B, N, W and Y................. 4-4 4-6. MAX GAP (LAMARS) Gap Criteria Summary .................................................... 4-5 4-7. MAX GAP (LAMARS) Correlation Confidence Levels ........................................ 4-8 4-8. MAX GAP (LAMARS) Curve Fit Correlation Values ........................................... 4-8 5-1. MAX GAP (VISTA) Case Characteristics (Witte and others, 2003)...................... 5-2 5-2. MAX GAP (VISTA) Pitch-to-Actuator Transfer Functions ( cG ) .......................... 5-2 5-3. MAX GAP (VISTA) Neal-Smith Pilot Models ...................................................... 5-3 5-4. MAX GAP (VISTA) Components for Cases B and W........................................... 5-4 5-5. Gap Criteria Values for MAX GAP (VISTA) Cases B and W ............................... 5-5 5-6. MAX GAP (VISTA) Components for Case N........................................................ 5-5 5-7. Gap Criteria Values for MAX GAP (VISTA) Case N ............................................ 5-5 5-8. Gap Criteria for MAX GAP (VISTA) Case Y ....................................................... 5-6 5-9. MAX GAP (VISTA) Gap Criteria Summary ......................................................... 5-6 5-10. MAX GAP (VISTA) Correlation Confidence Levels ........................................... 5-8 5-11. MAX GAP (VISTA) Curve Fit Correlation Values.............................................. 5-9 C-1. Minimum Values of the Correlation Coefficient for Confidence Level
(Wheeler and Ganji, 1996:147)............................................................................... C-2
xvii
List of Notations and Symbols
Symbol Definition Unit A Commanded Actuator Deflection deg Amax Maximum Actuator Deflection Available deg AFIT Air Force Institute of Technology - AFRL Air Force Research Laboratories - α, Alpha Angle-of-Attack deg dB Decibels - deg Degrees - δ Actuator Deflection Angle deg φ∆ Phase Angle rad
DoD Department of Defense - e, E Error - G(s) Generic Laplacian Transfer Function - Gactuator Actuator Transfer Function - Gaugmented Augmented Aircraft Transfer Function - GC Gap Criterion - Gc(s) Bare Aircraft Transfer Function - GP(s) Neal-Smith Pilot Model Transfer Function - HQDT Handling Qualities During Tracking - HUD Heads Up Display - jω Imaginary Laplacian Term -
xviii
Symbol Definition Unit k Feedback Gain Matrix - K* Intermediate Variable - Kp Neal-Smith Pilot Gain - Kα, Kalpha Angle-of-Attack Feedback Gain - Kq, Kq Pitch Rate Feedback Gain - LAMARS Large Amplitude Multimode Aerospace Simulator - Ln Natural Logarithm - MIL HDBK 1797 Military Handbook 1797 - n Number of Datapoints - N(A, ω) Describing Function - OLOP Open Loop Onset Point -
Phase Angle deg ω Frequency rad/sec ωa Actuator Gain - ωBW Band Width Frequency rad/sec ωsp Short Period Natural Frequency rad/sec PIO Pilot-induced Oscillations - PIOR Pilot-induced Oscillation Tendency Rating - q Aircraft Pitch Rate deg/sec R, rxy Correlation Factor - rad Radians - s Laplacian Operator -
xix
Symbol Definition Unit sec Seconds - θ, Theta Aircraft Pitch Angle deg T2 Time to Double Amplitude sec TLag Neal-Smith Lag Time Constant sec TLead Neal-Smith Lead Time Constant sec USAF TPS United States Air Force Test Pilot School - V True Velocity ft/sec VISTA Variable In-flight Stability Test Aircraft - VL Actuator Rate Limit deg/sec ζsp Short Period Damping Ratio -
1-1
AN INVESTIGATION RELATING LONGITUDINAL PILOT-INDUCED
OSCILLATION TENDENCY RATING TO DESCRIBING FUNCTION
PREDICTIONS FOR RATE-LIMITED ACTUATORS
I. Introduction
General
The purpose of this study was to investigate pilot-induced oscillations (PIO) and
determine a new method by which PIO tendency rating could be predicted. In particular,
longitudinal PIO in the presence of rate-limited actuators were singled out for
examination. The PIO rating scale used in this investigation was the PIO tendency
classification of the Department of Defense Interface Standard Flying Qualities of Piloted
Aircraft (MIL-HDBK-1797, 1997:152). While there are a number of PIO prediction
methods already published, this study will attempt a new approach.
This study was originated at the Air Force Institute of Technology (AFIT),
Wright-Patterson AFB, Ohio and was supported by the United States Air Force Test Pilot
School (USAF TPS). Research was conducted in both the Large Amplitude Multimode
Aerospace Simulator (LAMARS) at the Air Force Research Laboratory (AFRL), Wright-
Patterson AFB, Ohio and in the USAF NF-16D Variable In-flight Stability Test Aircraft
(VISTA) at Edwards AFB, CA. The VISTA aircraft is operated by USAF TPS and
supported by General Dynamics Advanced Information Systems of Buffalo, NY.
1-2
Background
Pilot-induced oscillations have been an aviation problem for over 100 years now.
The first incidence can be traced back to Wilbur and Orville Wright in 1903. When the
two brothers first took to the skies of Kitty Hawk, North Carolina, they experienced “a
mild longitudinal oscillation of the Wright Flyer” (Duda, 1995:288). The PIO problem
had just begun.
PIO Defined.
Before continuing with the century-long history of PIO, an understanding of the
term PIO is in order. A pilot-induced oscillation can be described as “an inadvertent,
sustained aircraft oscillation which is the consequence of an abnormal joint enterprise
between the aircraft and the pilot” (McRuer, 1995:2). Elaborated, a PIO is a complex
interaction between a pilot and his active involvement in an aircraft feedback system
(Klyde and others, 1995:14). The United State Department of Defense (DoD) defines
PIO as “sustained or uncontrollable oscillations resulting from the efforts of the pilot to
control the aircraft” (MIL-HDBK-1797, 1997:151). Although the key causal factor in
pilot-induced oscillation seems to be the pilot, it is important to make the assertion that,
generally, the pilot is not at fault and that there seems to be embedded in the flight control
system of the aircraft some tendencies predisposing the pilot-aircraft system toward PIO
occurrence (Klyde and others, 1995:14). In recent times, new terms have been put forth
to replace the familiar PIO such that the pilot’s guilt in such events is less likely to be
assumed. These include aircraft-pilot coupling (APC), pilot–in-the-loop oscillations and
pilot-assisted (or augmented) oscillations (Klyde and others, 1995:14). However,
experienced test pilots, including instructors at the US test pilot schools, and people in the
1-3
handling qualities community have expressed widespread allegiance to the traditional
term PIO and, therefore, this term will be used throughout this study (Mitchell and Hoh,
1995:16; Klyde and others, 1995:14).
In addition to defining what a PIO is, it is just as important to define what it is not.
A PIO could mean any oscillation that occurs during manual, piloted control. But some
of these situations could be the result of pilot overcontrol such as when a student pilot is
learning to land and balloons the aircraft. To an outsider, this could look like a PIO but
really is just part of standard pilot compensation lasting no more than one or two cycles
and is not a “real” PIO (Mitchell and Hoh 1995:17). Other researchers describe these as
“minor bobbles” that are often encountered as pilots get used to a new aircraft and is just
part of the learning experience (Klyde and others, 1995:14). It is also important to realize
that motions resulting from poor damping of the short period or dutch roll modes are not
PIO, when the motion does not result from the “efforts of the pilot to control the aircraft”
(Mitchell and Hoh, 1995:17-18).
To distinguish between these examples and a true PIO, some leading researchers
propose the following additional definition of PIO: “A PIO exists when the airplane
attitude, angular rate, or normal acceleration is 180 degrees out of phase with the pilot’s
control inputs” (Mitchell and Hoh, 1995:18). A great example of this phase lag can be
seen in Figure 1-1. This is the recorded data of the YF-22A accident which occurred on
25 April 1992 during a planned go-around at low altitude. This stripchart data depict a
180 degree phase difference between the aircraft pitch attitude and stick input.
The YF-22A PIO occurrence is just one of the most recent events. There is a long
history of PIO events in both operational and test flying as shown in Table 1-1.
Table 1-1. Summary of Famous Longitudinal PIO Events (McRuer, 1995:9)
Aircraft Type Summary of Incident XS-1 PIO during gliding approach and landing, 24 Oct 1947
XF-89A PIO during level off from dive recovery, early 1949 F-100 PIO during tight maneuvering X-15 Gliding flight approach and landing, 8 Jun 1959; Category II PIO
XF2Y-1 Post-takeoff destructive PIO YF-12 Mid-frequency severe PIO; Category III PIO
Space Shuttle ALT-5 during landing approach glide, 26 Oct 1977; Category II PIO DFBW F-8 PIO during touch and goes, 18 Apr 1978; Category III PIO
YF-22 PIO after touchdown and wave off in afterburner, 25 Apr 1992 JAS 39 PIO during approach, 1990; 1993; Category II – III PIO MD-11 China Eastern Airlines Flt 583, 6 Apr 1993; Inadvertant slat deployment
F-4 Low altitude record run second pass, 18 May 1961; Destructive PIO
1-5
The onset of PIO occurs when the pilot attempts tight control. The DoD defines
Category A Flight Phases as “those nonterminal Flight Phases that require rapid
maneuvering, precision tracking, or precise flight path control” (MIL-HDBK-1797,
1997:80). Types of maneuvers included in this category are in-flight refueling (receiver),
air-to-air combat, and formation flying (MIL-HDBK-1797, 1997:80). Category C Flight
Phases also require “accurate flight-path control” and include takeoffs, approaches and
landings (MIL-HDBK-1797, 1997:80). Almost all of the events listed in Table 1-1 fall
into Category A or Category C.
PIO can range in severity from benign to catastrophic (Anderson and Page,
1995:278). Benign PIO may not lead to loss of aircraft or life, but they can affect
operational missions. For example, receiver air-refueling may take longer than expected
causing delays in mission progress (Anderson and Page, 1995:278). This is still far easier
to accept than the more severe extreme in which aircraft and pilots have both been lost
due to PIO. On 18 May 1961, US Navy pilot Cmdr Jack Feldman and his F-4B aircraft
were lost when a PIO disintegrated his aircraft during the second pass of a low altitude
speed record attempt due to PIO related to low short-period damping (McCruer, 1995:9).
Even worse, the potential for PIO has actually increased due to the evolving use
of highly augmented, fly-by-wire aircraft (Liebst and others, 2002:740). Development
flight test of the Boeing 777, C-17A and Saab JAS 39 Gripen bear this out. The 777, a
fly-by-wire transport, encountered a PIO situation during a Flaps 20 landing. The
spoilers automatically deployed upon touchdown and the pilot countered with forward
column motion resulting in a PIO that lasted for “about 3 cycles” (Dornheim, 1995:32).
C-17 longitudinal PIO tendencies were discovered during dutch roll testing, air-refueling
1-6
breakaways and three-engine landings. Additionally, further PIO tendencies were
exposed during handling qualities during tracking (HQDT) exercises with an A-37 target
aircraft (Preston and others, 1996:20-49). The discovery of PIO tendencies in the JAS 39
was not so uneventful. Two aircraft were lost during different stages of aircraft
development. The first accident occurred in 1989 with the test aircraft having had only
5.3 hours of flight time when, during a landing attempt, it experienced “roll pendulums”
at low altitude and later “pendular movements in pitch, also” at an altitude of
approximately 10 to 15 meters. The mishap pilot attempted a go-around but the aircraft
“hit the runway with the left wingtip and main gear as well as the nose portion” (Ahlgren,
1989:12). The pilot survived and returned to test flying the JAS 39. In 1993 while
performing in an airshow in Stockholm, Sweden, he experienced a lateral PIO
compounded with an uncommanded pitch up to high angle-of-attack. The aircraft
diverged from controlled flight but the pilot ejected safely (Jensen and others,
undated:12).
PIO Categories.
PIO can be divided into three categories (McRuer, 1995:79-80). Category I PIO
are essentially linear pilot-vehicle system oscillations; they are usually the low frequency
consequence of excessive high frequency lag in the aircraft’s linear dynamics (Klyde and
others, 1996:17). On the other hand, Category III PIO are associated with essentially
nonlinear pilot-vehicle systems with transitions; they are the result of abrupt shifts in
either the effective controlled-element dynamics or in the pilot’s behavioral dynamics
(Klyde and others, 1996:17). Category II PIO are defined as quasi-linear pilot-vehicle
system oscillations associated with control surface rate or position limiting (McRuer,
1-7
1995:80). In other words, rate-lmited actuators. They represent a transition from linear
dynamics to nonlinear effects and were the focus of this investigation.
PIO Category II.
One reason for focusing on Category II PIO is that “almost all of the severe PIO
time history records available for operational and flight test aircraft … show surface rate
limiting …in the fully developed oscillation” (Klyde and others, 1996:15). Other
researchers have similar opinions: “It seems to be true that all recent PIO have exhibited
rate saturation” (Duda , 1995:289). A prime example is the development of the C-17 in
which “rate limiting was involved in all the subject [PIO] events and is viewed to be the
primary cause of the longitudinal PIO” (Preston and others, 1996:20).
There are two major detrimental effects of rate limiting. One is that it adds to the
effective lag in series with the pilot therefore making the effective aircraft dynamics
worse and the other is that it exposes the bare aircraft dynamics in stability augmented
aircraft. On the positive side, however, rate limiting tends to confine the ultimate
amplitude of the pilot-vehicle system oscillation (Klyde and others, 1996:15). Figure 1-2
shows the differences in the output of a typical first-order actuator as the effects of
increasing amplitude drive the system toward rate limiting. The figure displays the
characteristic triangular wave pattern of the actuator position reversing when equal to the
commanded position. The “boxcar-like” rate response is clipped at 40 deg/sec in this
example and is also a trait of a rate-limited actuator (Klyde and others, 1996:25). The
output was generated using the MatlabTM/SimulinkTM software model which is shown in
Figure 1-3 (MatlabTM/SimulinkTM, 2001).
1-8
Figure 1-2. Actuator Saturation Example Using an Input = A sin(ωi t)
Figure 1-3. SimulinkTM Actuator Model
PIO Scales.
Before discussing PIO prediction, a scale must be selected which distinguishes
increasing severity of PIO. Figure 1-4 shows just such a scale, the PIO tendency
classification (MIL-HDBK-1797, 1997:152). This scale is in common use among test
1-9
pilots to rate the susceptibility of aircraft to PIO and a large amount of historical data is
available based on this scale. Comparing this scale to Cooper-Harper Handling Qualities
Levels, a PIO rating of 1 or 2 is approximately equivalent to a Level 1 aircraft, a PIO
rating of 3 or 4 is similar to a Level 2 aircraft, a PIO rating of 5 is Level 3 and a PIO
rating of 6 is extremely dangerous (MIL-HDBK-1797, 1997:151). These PIO ratings are
more fully described in the PIO tendency rating scale of Table 1-2 (MIL-HDBK-1797,
1997:322).
Figure 1-4. PIO Tendency Classification (MIL-HDBK-1797, 1997:152)
1-10
Table 1-2. PIO Tendency Rating Scale (MIL-HDBK-1797, 1997:153)
PIO Prediction Methods.
Using this scale, several criteria for PIO prediction have been developed for
Category I and Category II PIO (Mitchell and Klyde, 1998:417-426). Category I PIO
prediction methods include the Bandwidth/Pitch Rate Overshoot Method, the original
Neal-Smith criteria, the Neal-Smith criteria as modified by the Moscow Aviation
Institute, Smith-Geddes PIO criteria, and the Gibson criteria (Mitchell and Klyde,
1998:417-426). A thorough examination of these methods led Mitchell and Klyde to
conclude that “we are in reasonably good shape in predicting Category I PIO” (Mitchell
and Klyde, 1998:417-426). Satisfactory development of Category II PIO prediction
methods, on the other hand, is not as complete.
DESCRIPTION NUMERICALRATING
No tendency for pilot to induce undesirable motions. 1 Undesirable motions tend to occur when pilot initiates abrupt maneuvers or attempts tight control. These motions can be prevented or eliminated by pilot technique.
2
Undesirable motions easily induced when pilot initiates abrupt maneuvers or attempts tight control. These motions can be prevented or eliminated but only at sacrifice to task performance or through considerable pilot attention and effort.
3
Oscillations tend to develop when pilot initiates abrupt maneuvers or attempts tight control. Pilot must reduce gain or abandon task to recover.
4
Divergent oscillations tend to develop when pilot initiates abrupt maneuvers or attempts tight control. Pilot must open loop by releasing or freezing the stick.
5
Disturbance or normal pilot control may cause divergent oscillation. Pilot must open control loop by releasing or freezing the stick. 6
1-11
Some work in Category II prediction methods has been accomplished or is in
progress. These efforts include modifying the Bandwidth/Phase Delay criteria to
Category II cases, the Time-domain Neal-Smith criteria, the Open Loop Onset Point
(OLOP) criteria and a power spectral density method using a structural model of the
human pilot (Mitchell and Klyde, 1998:417-426). These efforts continue despite the
assertion by Anderson and Page that “…the adaptive nature of the pilot makes such
oscillations [PIO] difficult to predict” (Anderson and Page, 1995:278).
Gap Criterion.
The Department of Aeronautical and Astronautical Engineering at the Air Force
Institute of Technology (AFIT) now proposes a new criterion for predicting PIO ratings
dubbed the Gap Criterion. It is based on describing function techniques, modified Neal-
Smith pilot models and actuator input amplitude ratios. The goal of this study was to
determine if specific relationships between Gap Criterion and PIO tendency rating for
Category II PIO due to rate-limited actuators exists and, if so, to what extent.
Objectives
The primary objective of this study was to develop a new criterion for predicting
PIO tendency rating by:
1) Exploiting previously defined describing function methods for determining
Category II PIO characteristics based on rate-limited actuators.
2) Defining Gap Criterion.
3) Applying the new method to existing historical data from similar test programs.
4) Conducting both simulator and flight tests to expand this database.
5) Making recommendations on the implementation of this new criterion.
1-12
Approach
The following steps were taken for this project:
1) The Gap Criterion was applied to data from previous simulator and flight test
projects in which rate limiting PIO effects were studied. These included USAF
TPS projects HAVE OLOP and HAVE PREVENT (Gilbreath, 2001; Hanley,
2003). The Gap Criterion was calculated for each test case of bare aircraft
dynamics and rate-limit. These Gap Criteria were then matched with their
respective PIO tendency ratings and plotted in pairs. Correlation probability
confidence level was then determined along with curve fits.
2) Based on the observed data from these studies, a broader range of longitudinal
flight control system dynamics with varying short period characteristics were
chosen to augment the database. These configurations were tested in both the
LAMARS simulator and the VISTA NF-16D aircraft.
Scope
This research project was limited in scope and constrained in certain areas:
1. The PIO investigated were strictly longitudinal, Category II PIO due to rate
limiting of the actuator.
2. Only three rate limits were chosen: 15 deg/sec, 30 deg/sec and 60 deg/sec.
3. Only four distinct bare aircraft dynamics cases were tested.
4. Tracking tasks were created with HUD generated and target aircraft sorties. No
offset landing tasks were planned.
5. Due to schedule and budget, the simulator test portion was limited to two days and
three pilots. The flight test portion was limited to eight sorties and 10.8 hour
2-1
II. Theory
This chapter will discuss describing functions and how they can be used to
understand and predict PIO onset when considered in the context of rate-limited
actuators. Further, the Neal-Smith pilot model will be explained followed by an example
integrating all of these concepts. The basis of the Gap Criterion will then be covered.
Describing Function Development
Observing the time history of the YF-22 PIO from Figure 1-1, it can be seen that
the input is approximately sinusoidal. This is true in general of all PIO incidents (Klyde
and others, 1996:37). The describing function technique can be used for limit cycle
analysis due to the fact that that the form of the signals in a limit-cycling system, such as
a PIO, is usually approximately sinusoidal (Slotine and Li, 1991:157).
Any system which can be rearranged into the form shown in Figure 2-1, where w
and G(p) represent nonlinear and linear elements, respectively, can be studied using
describing functions (Slotine and Li, 1991:162). Examples of nonlinear elements include
dead-zones, hysteresis or rate saturations. Rate saturations were the focus of this study.
w=f(x) G(p) + - y(t) r(t) = 0 x(t) w(t)
Figure 2-1. Example of a Nonlinear System
For the basic version of the describing function method, the system has to satisfy
the following four conditions (Slotine and Li, 1991:164):
2-2
1) There is only a single nonlinear component
2) The nonlinear component is time-invariant
3) Corresponding to a sinusoidal input sin( )x A tω= , only the fundamental
component w1(t) in the output w(t) has to be considered
4) The nonlinearity is odd
Consider a sinusoidal input of the form sin( )x A tω= entering the nonlinear
element of the system shown in Figure 2-1. Due to nonlinear effects, the output, w(t), is
“often a periodic though non-sinusoidal function” (Slotine and Li, 1991:165). The output
function w(t) can be expanded using Fourier series as seen in Equation 1 and the
succeeding derivation (Slotine and Li, 1991:165):
0
1( ) [ cos( ) sin( )]
2 n nn
aw t a n t b n tω ω∞
=
= + +∑ (1)
where
01 ( ) ( )a w t d t
π
π
ωπ −
= ∫ (2)
1 ( )cos( ) ( )na w t n t d tπ
π
ω ωπ −
= ∫ (3)
1 ( )sin( ) ( )nb w t n t d tπ
π
ω ωπ −
= ∫ (4)
Applying condition four above, for all odd functions 0 0a = (Slotine and Li,
1991:166). Further, applying the third assumption means discarding all other terms
except 1n = (Slotine and Li, 1991:166). This leaves:
1 1 1( ) ( ) cos( ) sin( )w t w t a t b tω ω≈ = + (5)
2-3
which can be rewritten as
1( ) sin( )w t M tω φ= + (6)
where
2 21 1M a b= + (7)
1 1
1
tan ab
φ − =
(8)
Rewritten in complex notation leads to:
( ) ( )1 1 1( ) ( )j t j tw t Me b ja eω φ ω+= = + (9)
Finally, the describing function, ( , )N A ω , is defined to be the complex ratio of
the fundamental component of the nonlinear element to the input sinusoid. This is shown
in Equation 10:
( )( )
1 1( )
1( , )j t
jj t
Me MN A e b jaAe A A
ω φφ
ωω+
= = = + (10)
Saturation Nonlinearity Describing Function
Now consider the saturation input-output relationship shown in Figure 2-2 below:
γ ka
ka 0 ωt
w(t) w
0
k
ωt
γ π/2
A x(t) 0
a x
Figure 2-2. Saturation Nonlinearity and the Corresponding Input-Output Relationship
(Slotine and Li, 1995:173)
2-4
From the figure, it is apparent that if our input, ( ) sin( )x t A tω= , has a maximum
amplitude A ≤ a then the input remains linear and the output is just w(t) = kAsin(ωt). But
if the maximum amplitude, A, is greater than a, clipping occurs and the value of w(t) can
be split into two sets over the first quarter of the symmetric output (Slotine and Li,
1995:173):
sin( )
( )kA t
w tka
ω=
0
2
t
t
ω γπγ ω
≤ ≤
< ≤ (11)
where γ = sin-1(a/A)
The output w(t) is an odd function, implying 1 0a = in Equation 5. Further,
dividing the output into four quarters yields a new equation for b1
2
10
4 ( )sin( ) ( )b w t t d tπ
ω ωπ
= ∫ (12)
2
21
0
4 4sin ( ) ( ) sin( ) ( )b kA t d t ka t d tπγ
γ
ω ω ω ωπ π
= +∫ ∫ (13)
2
1 2
2 1kA a abA A
γπ
= + −
(14)
Substituting 1 0a = and Equation 14 into Equation 10 leaves (Slotine and Li, 1995:174):
2
112
2( , ) sin 1b k a a aN AA A A A
ωπ
−
= = + −
(15)
Closed Loop Describing Function Approximation
Now, consider the block diagram in Figure 2-3 of a first order actuator system and
the derivations which follow (Klyde and others, 1996:36-46).
2-5
eδ
e s1eδ eδe
+−
LV
Le
aω
LV-
1
commandeδ
Figure 2-3. Actuator Model Development (Klyde and others, 1995:22)
The nonlinear portion of this model is exactly the same as the saturation
nonlinearity discussed previously. Substituting the appropriate new nomenclature and
letting ( ) sin( )e t E tω φ= + replace x(t), leads to the following describing function for the
nonlinear element:
2
12
2( , ) sin 1a L L Le e eN AE E E
ωωπ
− = + −
(16)
Further, by using series expansions for both the arcsine term and the square root, the
describing function can be approximated by:
3 2
2
2 1 1( , ) 16 2
a L L L Le e e eN AE E E E
ωωπ
= + + ⋅⋅⋅ + − − ⋅⋅⋅ (17)
Keeping only the first order linear terms yields:
2 4( , ) a aL L Le e eN AE E E
ω ωωπ π
= + = (18)
Substituting L a LV eω= leads to:
4( , ) LVN AE
ωπ
= (19)
2-6
Next, consider the revised block diagram shown in Figure 2-4 and determine the
closed loop transfer function, treating N as a constant.
N 1s + -
δ (s) e
δ (s) e Command e(s)
Figure 2-4. Closed Loop Actuator Transfer Function Diagram
Treating N as a constant and utilizing linear block diagram transfer function
techniques, the relationship of e(s) to ( )Commande sδ is:
( ) 1( ) 1Commande
e sNss
δ=
+ (20)
Assuming ( ) sin( )Commande t A tδ ω= and ( ) sin( )e t E tω φ= + and substituting jω for s,
the equation for the magnitude of this transfer function becomes:
2
2
( ) 1 sin( )( ) sin( )
1Commande
e s E t Es A t AN
ω φδ ω
ω
+= = =
+
(21)
Rearranging Equation 19 in terms of E gives:
4 LVENπ
= (22)
Substituting Equation 22 into Equation 21 and rearranging terms yields:
2
14 L
NAV
ω
π ω=
−
(23)
2-7
Now, still treating N as a constant and utilizing standard block diagram transfer
function techniques, the relationship of ( )e sδ to ( )Commande sδ is:
( ) 1( ) 1Command
e
e
s Nss s NN
δδ
= =++
(24)
and its magnitude is
2
2 2
( )( )
Command
e
e
j Nj N
δ ωδ ω ω
=+
(25)
and substituting Equation 23 into Equation 25 gives
( ) 4( )
Command
e L
e
j Vj A
δ ωδ ω π ω
= (26)
Solving for the phase angle of Equation 24 yields:
1( ) tan( )
Command
e
e
jj N
δ ω ωδ ω
− − =
(27)
and substituting Equation 23 yields
2
1( ) tan 1( ) 4
Command
e
e L
j Aj V
δ ω π ωδ ω
− = − −
(28)
Sinusoidal Input/Triangle Output Describing Function Approximation
Another describing function approximation can be made by utilizing the observed
characteristics of a saturated actuator. The input, xi(t), is sinusoidal in nature and the
output, x0(t), takes on the familiar saw tooth triangle shape as shown in Fig 2-5 (Klyde
and others, 1995:42-46):
2-8
t0
t i
x 0
x i x i (t)
tD
x0(t)
Rate LimitingElement
x0(t) x i (t)
Figure 2-5. Rate-Limiting Input and Output (Klyde and others, 1995:42-46)
As before, let the input be sinusoidal as shown in Equation 29:
max
( ) sin( )i ix t x tω= (29)
and the derivative or input rate is:
max
( ) cos( )i ix t x tω ω= (30)
Now, let 2 Tω π= where 4 iT t= . Then the maximum input rate is
max
max 2i
ii
xx
tπ
= (31)
The rate of the output, 0x , is equal to the slope of the output and is given by
00
0
xxt
= ± (32)
Now, take the relationship of the output rate to the input rate in the range of t0 and
solve for the ratio of output to input magnitude as:
2-9
max
maxmax
0 0 0
0 0
22
i i
i ii
xx x x tt t t xx
ππ
= =
(33)
Recognizing t0 equals ti, rearranging terms and introducing a new variable *K gives:
max max
0 0 *2i i
x x Kx x
π= = (34)
Rewriting this expression in terms of the Figure 2-3 variables and recognizing that the
output rate when saturated is VL and the maximum input rate is Aω leaves
*2
LVKA
πω
= (35)
The describing function magnitude is then expressed using the *K value multiplied by
the Fourier fundamental of the triangle wave as seen in Equation 36 (Klyde and others,
1996:45).
2
( ) 8 4*( )
Command
e L
e
j VKj A
δ ωδ ω π π ω
= = (36)
This is exactly the same expression derived earlier for the closed loop actuator describing
function magnitude. To obtain the phase angle of the input/output relationship, the term
tD as shown in Figure 2-5 must be determined. The input and output amplitudes are equal
when i Dt t t= + .
( )max 0sini i Dx t t xω + = (37)
Simplifying this expression by substituting max0* iK x x= , expanding sin[ω(ti + tD)], and
substituting 2itω π= results in (Klyde and others, 1996:45):
cos( ) *Kφ∆ = (38)
2-10
where Dtφ ω∆ = is the phase angle between the input and output. Solving for φ∆ and
noting that it is a phase lag gives Equation 39:
2
1 1( ) 1cos ( *) tan 1( ) *
Command
e
e
jKj K
δ ωφδ ω
− − −∆ = − = = − −
(39)
Now to compare with the closed-loop describing function phase angle, substitute
( )( )* 2 LK V Aπ ω= into Equation 39, which results in
2
1( ) 2tan 1( )
Command
e
e L
j Aj V
δ ω ωδ ω π
− = − −
(40)
This is slightly different from the closed loop describing function phase angle expression.
These phase angle differences are shown as a function of *K in Figure 2-6:
Figure 2-6. Describing Function Phase Angle Comparison
2-11
These two describing function approximations were introduced to show that
similar results can be derived from different methods. According to Klyde, the more
accurate of these two describing function approximations for application to Category II
PIO is the sinusoidal input/triangle output solution (Klyde and others, 1996:46).
Therefore, this describing function will be used throughout the remainder of this study.
Applying Describing Function Results to Predict PIO
Consider the longitudinal closed-loop system shown in Figure 2-7. ( )pG s
represents a model of the pilot and Gc (s) represents a model of the bare aircraft
dynamics. The remaining elements are equivalent to the rate-limited actuator model
previously discussed in Figure 2-3.
G p (s) G c (s) 1 s
e(s) θ Command (s) θ Error (s) θ (s) δe(s)δ e Command (s) δe(s)
.VL
-VL
ωa
k
Figure 2-7. Pitch Tracking Closed Loop System
The linear elements ( )pG s and ( )cG s can be combined into one linear element,
( )G s and the nonlinear element, ( , )N A ω , remains separate as shown in Figure 2-8.
N(A,ω) G(s)
+ -
θ Command (s)
θ Error (s)
θ (s)
Figure 2-8. Simplified Pitch Tracking Closed Loop System
2-12
This model can then be applied to a limit cycle analysis. The requirement for a
neutrally damped oscillation is simply that the open-loop amplitude ratio be equal to 1.0
and the phase be -180º (Klyde and others, 1996:54). In order for a PIO to persist, the
system shown in Fig 2-8 must satisfy the Nyquist criteria shown in the following
equation (Klyde and others, 1996:54):
( ) ( , ) 1G j N Aω ω = − (41)
or rearranged
1( )( , )
G jN A
ωω−
= (42)
The easiest way to view the application of this equation is to plot the open-loop
magnitude and phase values of the negative inverse describing function, 1 ( , )N j Aω− ,
using the *K solutions from Equations 36 and 39 as well as the open-loop magnitude
and phase of ( )G jω . If the two plots intersect, a PIO is predicted (Klyde and others,
1996:63). This will be shown by means of an example later in this chapter. The *K
solutions for the negative inverse describing function are shown below in Equations 43
and 44 (Liebst, 2002):
10 2
1 8 *20( *)
KLogN K π− = −
(dB) (43)
( )11 180 cos * 180( *)
KN K π
−−= − (deg) (44)
2-13
Pilot Model
There are many pilot models to choose from in the literature. Some believe that a
simple gain with no phase lag best represents the pilot in the PIO situation (Klyde and
others, 1996:54). Others believe structural models are better predictors (Mitchell and
Klyde, 1998:426). In another recent study, the Neal-Smith pilot model was judged to
best represent the pilot model prior to the onset of rate limiting (Gilbreath, 2001:7-3).
Therefore, in this study, the Neal-Smith pilot model will be utilized.
The Neal-Smith pilot model is useful for pilot-aircraft pitch attitude control loops
with unity-feedback and has the following characteristics (MIL-HDBK-1797, 1997:237):
1. Adjustable gain
2. Time delay
3. Ability to develop lead, or to operate on derivative or rate information
4. Ability to develop lag, or to smooth inputs
5. Ability to provide low-frequency integration
The Neal-Smith pilot model can take on one of two forms. This determination is
based on the whether constant speed or two-degree-of-freedom equations are used to
represent the bare aircraft dynamics. These are typified by noting whether or not a free
integrator is contained in the denominator of the aircraft pitch transfer function.
Otherwise, three-degree-of-freedom equations or flight control system utilizing attitude
stabilization will require a different form. Table 2-1 shows these two transfer functions
for the Neal-Smith pilot model (MIL-HDBK-1797, 1997:237; Bailey and BidLack,
1995:8).
2-14
Table 2-1. Neal-Smith Pilot Models Aircraft Transfer Function
with a Free Integrator Aircraft Transfer Function without a Free Integrator
( )( )
0.251( )
1Lead s
p pLag
T sG s K e
T s−+
=+
( ) ( )
( )0.255 1 1
( )1
Lead sp p
Lag
s T sG s K e
s T s−+ +
=+
The theory states that the pilot chooses his gain, pK , and his lead/lag time
constants, TLead and TLag, to attain a certain bandwidth. This bandwidth varies with the
flight phase category. For example, for Category A flight phase maneuvers such as air-
to-air dogfighting, the required bandwidth is 3.5 rad/sec. This is measured at a closed-
loop phase of –90 degrees. Further, the pilot adjusts pK , TLead and TLag to minimize droop
to no greater than 3 dB for Level 1 performance and no greater than 9 dB for Level 2
over the frequency range from 0 to 10 rad/sec while at the same time minimizing closed
loop resonance (MIL-HDBK-1797, 1997:239). The phase lag term, e-0.25s, represents
delays in the pilot’s neuromuscular system (MIL-HDBK-1797, 1997:239). A graphical
depiction of these pilot efforts can be seen in Figure 2-9.
2-15
Figure 2-9. Neal-Smith Pilot Model Constraints
Max Droop < 3 dB for Level 1
Minimize Resonance
Closed Loop Bode Diagram
( )( )
ssCommand
θθ
( )( )
ssCommand
θθ
BWω
2-16
Gap Criterion
Utilizing the previous theoretical developments, a systematic process relating
aircraft plant dynamics and actuator rate limits to PIO tendency rating will be introduced.
The procedure is called the Gap Criterion and is based on the block diagram shown in
Figure 2-10:
G p (s) G c (s) 1 s
e(s) θ Command (s) θ Error (s) θ (s) δe(s)δ e Command (s) δe(s)
.VL
-VL
ωa
k
*actuator augmentedG G
Figure 2-10. Pitch Tracking Closed-Loop System for Gap Criterion
In modern fly-by-wire aircraft, feedback is an integral part of obtaining more
desirable closed loop flying qualities. However, as mentioned earlier, rate limiting
exposes the unaugmented dynamics and adds phase lag (Hanley, 2003:1-3). A pilot
suddenly faced with different flying qualities will not be able to adjust his gain, lead or
lag properties instantaneously. He will therefore continue to fly in such a manner as if
the augmented aircraft dynamics were still in place. Therefore, the term G(jω) from
Equation 42 is the product of the bare aircraft dynamics, Gc(s), convolved with the Neal-
Smith pilot model, Gp(s). This idea is incorporated in the derivation of the Gap
Criterion.
When rate limiting is not occurring, the actuator dynamics from Figure 2-10 can
be determined from block diagram methods and is Gactuator = ( )a asω ω+ .
2-17
Gap Criterion Formulation.
Computing the Gap Criterion consists of the following steps:
1. Determine the bare aircraft pitch-to-actuator transfer function, ( )( )( )c
e
sG ss
θδ
= .
2. If the short period poles of ( )cG s are unstable then the Gap Criterion
automatically equals zero. This is due to control amplitudes approaching
zero which cause an immediate departure from controlled flight due to
dynamic instability resulting from actuator rate saturation.
3. Determine actuator dynamics for the following form: aactuator
a
Gsωω
=+
.
Typically, 20aω = and this will be used throughout this study (Liebst,2001).
4. Determine an appropriate optimized Neal-Smith pilot model, ( )pG s , for the
augmented aircraft dynamics plus actuator dynamics, *actuator augmentedG G .
5. Plot the open-loop magnitude and phase of the bare aircraft dynamics
convolved with the Neal-Smith pilot model dynamics, ( )* ( )c pG s G s on a
Nichols chart
6. Plot the negative inverse describing function open-loop magnitude and phase
on the same Nichols chart using *K . See Equations 43 and 44.
7. Determine the resulting type by reference to Figure 2-11 and then compute
the Gap Criterion by following the steps of that type.
2-18
Type I Type II
Type III Type IV
Figure 2-11. Four Resulting Gap Criterion Types
2-19
Type I.
1-1. Determine the minimum amount by which the pilot would need to
effectively increase gain, pK∆ (dB), such that the two magnitude-phase
lines just intersect at a frequency greater than the -3 dB Neal-Smith
maximum droop frequency as shown in Figure 2-12:
Figure 2-12. Case I Effective Pilot Gain Increase
1-2. Determine the values of K* and ω (rad/sec) at this intersection
1-3. Determine the commanded actuator deflection amplitude, A, utilizing the
following equation where VL is the known actuator rate limit in deg/sec:
2 *LVA
Kπω
=
1-4. Normalize this amplitude by dividing by the maximum available actuator
deflection, Amax
1-5. The Gap Criterion is this normalized amplitude multiplied by ∆Kp:
( / 20)
max
*10 pKAGap CriterionA
∆=
K p∆
2-20
Type II.
2-1. Determine the minimum amount by which the pilot would need to
effectively decrease gain, pK∆ (dB), such that the two magnitude-phase
lines only intersect in one place at a frequency greater than the -3 dB Neal-
Smith maximum droop frequency as shown in Figure 2-13:
Figure 2-13. Case II Effective Pilot Gain Decrease
2-2. Determine the values of K* and ω (rad/sec) at this intersection
2-3. Determine the commanded actuator deflection amplitude, A, utilizing the
following equation where VL is the known actuator rate limit in deg/sec:
2 *LVA
Kπω
=
2-4. Normalize this amplitude by dividing by the maximum available actuator
deflection, Amax
2-5. The Gap Criterion is this normalized amplitude multiplied by ∆Kp.
( / 20)
max
*10 pKAGap CriterionA
∆=
K p∆
2-21
Type III.
3-1. Determine the values of K* and ω (rad/sec) at the intersection
3-2. Determine the commanded actuator deflection amplitude, A, utilizing the
following equation where VL is the known actuator rate limit in deg/sec:
2 *LVA
Kπω
=
3-3. Normalize this amplitude by dividing by the maximum available actuator
deflection, Amax
3-4. The Gap Criterion is this normalized amplitude.
max
AGap CriterionA
=
Type IV.
4-1. No determination of Gap Criterion can be made.
Example of Gap Criterion Application
Reconsider the closed loop system of Figure 2-10 with the following characteristics:
• ( )( )( )2
4.5 1.5( )
3 6c
sG s
s s s+
=+ +
• 2020actuatorG
s=
+
• k = 0 (unaugmented), therefore augmented cG G=
• VL = 30 deg/sec
• Maximum actuator deflection: max
30 degeδ =
• Category A flight phase, Neal-Smith bandwidth = 3.5 rad/sec
2-22
Utilizing the USAF Wright Laboratory Flight Dynamics Directorate’s MATLABTM
Interactive Flying Qualities Toolbox for Matlab (Domon and Foringer, 1996), the Neal-
Smith pilot model was found to be:
0.25sp e
1)(0.0001s1)(0.583s0.856(s)G −
++
=
The open-loop magnitude and phase of *actuator augmentedG G are plotted in Figure 2-14 as
well as the open-loop magnitude and phase of the negative inverse describing function. It
can be seen that an open-loop gain increase (∆Kp) of 7.502 dB is required for the two
lines to meet. At this intersection, the values for K* and ω are 0.7635 and 3.9418 rad/sec,
respectively. After calculating the amplitude, A = 15.66 deg, the result is normalized by
dividing by 30 deg (δe max). This normalized result is multiplied by ∆Kp to yield the Gap
Criterion. In this example the Gap Criterion equals 1.238.
This example showed how preflight Gap Criterion can be calculated. It is
expected that matching these values with their respective PIO Tendency Ratings should
yield some correlation. This will be accomplished by examining historical databases and
then collecting additional data in the LAMARS simulator and VISTA aircraft.
Figure 2-14. Nichols Chart of the Example Problem
∆Kp = 7.502 dB
K* = 0.7635 ω = 3.9418 rad/sec
3-1
III. Analysis of Selected Historical Data
In this section, two previous PIO studies, HAVE PREVENT and HAVE OLOP,
will be examined using the Gap Criterion (Hanley, 2002; Gilbreath, 2000).
HAVE PREVENT Analysis
HAVE PREVENT was a simulator and flight test study comparing two different
PIO prevention filters. The study was conducted as part of a Test Management Project at
the USAF Test Pilot School and a thesis sponsored by the Air Force Institute of
Technology (Hanley and others, 2002; Hanley, 2003). The two PIO filters examined
were the Feedback with Bypass and the Derivative Switching filters. In order to establish
a baseline, runs with neither filter engaged were conducted in both the LAMARS
simulator and VISTA NF-16D aircraft. However, few no-filter data points were collected
during the flight test portion. Therefore, only the simulator data will be examined.
In this simulator, pilot-induced oscillation tendency ratings (PIOR) were gathered
in two distinct phases with specific piloting tasks and in which the evaluation pilot was
blind to the randomized bare aircraft dynamics and rate limit combinations. In Handling
Qualities Phase 2 testing, a precision aimpoint was tracked as “aggressively and
assiduously as possible, always striving to correct even the smallest errors” using a
piloting technique known as Handling Qualities During Tracking (HQDT) (Brown and
others, 2002:21-18). The precision aimpoint involved a sum-of-sines head’s up display
(HUD) tracking task, as shown in Figure 3-1 (MIL-HDBK-1797, 1997:108m). Handling
Qualities Phase 3 “operational” testing was also accomplished and involved two different
tasks (Brown and others, 2002:21-19). These tasks were the discrete HUD pitch tracking
task and computer generated aircraft target tracking task shown in Figures 3-2 and 3-3.
HAVE PREVENT LAMARS DataMAX GAP LAMARS DataHAVE OLOP VISTA DataMAX GAPVISTA Data
2
Figure 6-7. LAMARS and VISTA Combined Phase 3 Discrete Pitch-Tracking Data
6-5
Combined Dataset Analysis and Observations
Each combined dataset far exceeded a 95% confidence level that correlation
existed. In fact, the lowest correlation confidence level was 99.95% for the LAMARS
Combined Phase 3 Target Tracking. The logarithmic curve fits shown in the previous
figures proved to be the best combination of high correlation factor and endpoint
matching for both high and low Gap Criterion values.
Further inspection of the plots reveals some trends. From the PIO tendency scale
shown in Figure 1-4 of Chapter I, a PIOR of 4, 5 or 6 represents a tendency for PIO while
a PIOR of 1, 2 or3 represents no tendency for PIO, though some undesirable motions are
still possible. From the combined dataset figures in can be seen that PIOR rating is an
inverse function of the Gap Criterion: High PIOR come at low Gap Criterion values and
vice versa. Also, from these figures it can be seen that, in general, the majority of PIO
tendency ratings 4, 5, and 6 occur at Gap Criterion values less than approximately 1.0
while PIO tendency ratings at Gap Criterion values above 1.0 tend to be non-PIO values
of 1, 2 or 3. These two observations were more pronounced in the Phase 2 Sum-of-sines
Tracking Task than in the Phase 3 tasks. This is possibly due to the high gain, high
bandwidth piloting technique used in this task while the other tasks were more focused on
task performances which required both high and low gains and bandwidths.
The Gap Criterion value of 1.0 also has physical significance. Consider a Type I
solution in which G(s) was such that ∆Kp approached zero. Then the Gap Criterion
would simply be the amplitude to cause PIO divided by the maximum amplitude
available and if this Gap Criterion equals 1.0 then the amplitude to cause PIO would be
equal to the maximum amplitude available. Hence, any Gap Criterion value greater than
6-6
1.0 would require more actuator deflection than was available, thereby providing a
natural limit to creating a PIO. In other words, if an amplitude of 35 degrees were
required, with only 30 degrees available, it is readily apparent that a PIO cannot be
achieved.
The relatively high correlation factors, especially for the Combined Phase 2 Sum-
of-sine datasets indicate the potential for Gap Criterion validation and acceptance as a
tool for predicting Category II PIO due to rate-limited actuators.
7-1
VII. Conclusions
In this study, a new criterion for predicting pilot-induced oscillation tendency
rating due to rate-limited actuators was developed and correlated to datasets of these
ratings. This criterion was called the Gap Criterion.
Two historical databases, Projects HAVE PREVENT and HAVE OLOP were
selected to see if the Gap Criterion had merit. Most datasets were assessed with greater
than 95% confidence that a correlation indeed existed. Further, a logarithmic curve fit
was deemed best. Follow on testing in Project MAX GAP gathered more PIO tendency
rating data to augment these earlier findings. These datasets also showed strong evidence
of correlation and again found a logarithmic interpolation of the data worked well.
These datasets were combined for different tasks and sources to determine
whether the Gap Criterion was universal in nature. This seemed to be the case with all
combined datasets indicating greater than 99.95% confidence level that correlation
existed between the PIOR and the Gap Criterion. Logarithmic curve fits again appeared
superior with high relative correlation factors and good endpoint matching.
Further observations were made for the combined datasets. Based on the relative
positions of a majority of the data, it was found that lower Gap Criterion values resulted
in higher PIO tendency ratings and vice versa. Further, proper PIO represented by PIO
tendency rating values of 4, 5 or 6 were clustered at Gap Criterion values of 1.0 or less
while PIO tendency ratings of 1, 2 or 3, representing non-PIO, were more prevalent at
Gap Criterion values greater than 1.0.
The Gap Criterion has merit. It should be used as a tool to predict and reduce
incidents of Category II PIO due to rate-limited actuators during aircraft development.
A-1
Appendix A. Matlab/SimulinkTM Code
This appendix lists the MatlabTM source code used to move the short period poles
of the F-16 to desired locations and then determine the feedback gains for angle-of-attack
and pitch rate necessary to augment the new bare aircraft dynamics and return them to
suitable closed loop dynamics. A SimulinkTM diagram is also shown and was used for
the preceding process. The next code listing computes the Gap Criterion for various bare
aircraft dynamics and rate limit choices.
Bare Aircraft Pole Placement
%Bare Aircraft Dynamics Bare Pole placement %This matlab file will take the bare F-16 dynamics at 15,000 ft Pressure Altitude, %300 KCAS and place the short period poles where I want them based %on what I give for short period damping and natural frequency. clear;clc;format short g; format compact warning off a = [-0.033104 0.14957 -0.3207 -0.56111; -0.015511 -1.2826 1 -0.0024621; 0.008081 -4.0875 -1.7556 0.0012828; 0 0 1 0] b = [-0.5193; -0.05243; -11.085; 0] c = eye(4) d = [0; 0; 0; 0] %Input the Desired Short Period Damping Ratio and Short Period Natural Frequency for test case %Example: MAX GAP LAMARS Case B zetasp = 0.61, omegansp = 2.34 zetasp = input('Short Period Damping Ratio:'); omegansp= input('Short Period Natural Frequency:'); sigma = zetasp*omegansp omegad = omegansp*sqrt(1-zetasp^2) %Compute P vector to "place" the poles for the desired Short Period Damping %Ratio and Short Period Natural Frequency. %Also, Phugoid pole locations are chosen as -0.017+/-0.074j %(Phugoid Damping Ratio = 0.224, Phugoid Natural Frequency = 0.075928 rad/s) P=[-.017+.074*j -.017-.074*j -sigma+omegad*j -sigma-omegad*j] kinner=-1*place(a,b,P) 'ahat is a+b*kinner'
A-2
ahat=a+b*kinner 'eigenvalues of ahat' eigenvalues = eig(ahat) 'natural frequencies and dampings of ahat' [wwn,zz]=damp(ahat) '(Phugoid Damping Ratio = 0.2239, Phugoid Natural Frequency = 0.075928 rad/s)' [num,den]=ss2tf(ahat,b,c,d) BareAcftLongDynamics=tf(num(4,:),den) continuing = input('continue'); %The following are the phugoid and short perios (sp1 & sp2) poles for the DESIRED %(must be changed if desired closed loop characteristics are different) %closed loop system using angle-of-attack (alpha) and pitch rate (q) feedback: phugoid1=-.017+.074*j;phugoid2=-.017-.074*j;sp1=-2.2+2.22*j;sp2=-2.2-2.22*j P2=[phugoid1 phugoid2 sp1 sp2] K=place(ahat,b,P2) %Find approximate values of Kq (call it Kqbase) and %Kalpha (call it Kalphabase) Kqbase=-K(1,3) Kalphabase=-K(1,2) %Find best Kq and Kalpha %This is an iterative, graphical technique to find Kalpha and Kq that take the %bare aircraft dynamics and change them into the desired closed loop dynamics %The error function {E(ii)=abs(wn(3)-3.125)+4.439*abs(z(3)-.7)} is important %in that it relates to the closed loop short period natural frequency desired (3.125) %and the short period damping ratio desired (.7). The constant 4.439 is a weighting funtion %so that frequency and damping ratio are considered equally (3.125/.7 = 4.439) %This is to get a ballpark Kalpha and Kq ii=1 for kq = Kqbase-1:.025:Kqbase+1; for kalpha = Kalphabase-1:.025:Kalphabase+1; Kqcounter(ii)= kq;Kacounter(ii)=kalpha; [A2,B2,C2,D2]=LINMOD('AugmentedDynamics',0); Yc=ss(A2,B2,C2,D2); [Yctfnum,Yctfden]=ss2tf(A2,B2,C2,D2); yctf=tf(Yctfnum,Yctfden); [wn,z]=damp(yctf); wnn(ii)=wn(3);zz(ii)=z(3);iii(ii)=ii; E(ii)=abs(wn(3)-3.125)+4.439*abs(z(3)-.7); ii=ii+1 end end plot(iii,E) grid on grid minor
A-3
xx=input('what index?') 'kq',Kqcounter(xx) 'kalpha',Kacounter(xx) Kqbase=Kqcounter(xx); Kalphabase=Kacounter(xx); %This is to get a more precise Kalpha and Kq ii=1 for kq = Kqbase-.1:.0025:Kqbase+.1; for kalpha = Kalphabase-.1:.0025:Kalphabase+.1; Kqcounter(ii)= kq;Kacounter(ii)=kalpha; [A2,B2,C2,D2]=LINMOD('AugmentedDynamics',0); Yc=ss(A2,B2,C2,D2); [Yctfnum,Yctfden]=ss2tf(A2,B2,C2,D2); yctf=tf(Yctfnum,Yctfden); [wn,z]=damp(yctf); wnn(ii)=wn(3);zz(ii)=z(3);iii(ii)=ii; E(ii)=abs(wn(3)-3.125)+4.439*abs(z(3)-.7); ii=ii+1 end end plot(iii,E) grid on grid minor xx=input('what FINAL index?') %Final results kq=Kqcounter(xx) kalpha=Kacounter(xx) [A2,B2,C2,D2]=LINMOD('AugmentedDynamics',0); Yss=ss(A2,B2,C2,D2); [Ynum,Yden]=ss2tf(A2,B2,C2,D2) Ytf=tf(Ynum,Yden) damp(Ytf)
A-4
Figure A-1. Augmented Dynamics SimulinkTM Model
Gap Criterion Computation
%This Matlab file will be used to take LAMARS Cases B, N, W and Y and compute the Gap Criterion % Maj Joel Witte % 04 Nov 03 %This Matlab file takes the bare aircraft dynamics (Gc) and multiplies by the closed loop modified %Neal-Smith pilot model (Gp). Then the delta gain, gap (or ∆Kp), is calculated and kstar and omega %are determined. %Finally, the Gap Criterion is determined for each max actuator rate. clear;clc;clf;format compact figure(1) % %The following are the parameters for the MAX GAP LAMARS evalutions. % Gc = bare aircraft dynamics %Gp = Neal-smith pilot model %G = bare aircraft dynamics convolved with the Neal-Smith Pilot Model %Choose which case by commenting/uncommenting the parameters %For example, currently Case B will be computed %Case B Gcb=tf([-11.08 -14.37 -0.5277],[1 2.889 5.578 0.2026 0.03157]) Gpb= -.11483*tf([.32483 1],[.00001 1])*tf([5 1],[1 0],'iodelay',.25); G=Gcb*Gpb %Case N % Gcn=tf([ -5.551e-015 -11.09 -14.37 -0.5277],[1 1.912 9.867 0.3439 0.05648]) % Gpn= -.10919*tf([.32699 1],[.00001 1])*tf([5 1],[1 0],'iodelay',.25); % G=Gcn*Gpn %Case W % Gcw=tf([-3.553e-015 -11.09 -14.37 -0.5277],[1 8.512 22.48 0.8031 0.1279]) % Gpw= -.15254*tf([.30311 1],[.00001 1])*tf([5 1],[1 0],'iodelay',.25); % G=Gcw*Gpw
A-5
%Case Y %Gcy=tf([-6.217e-015 -11.08 -14.37 -0.5277],[1 4.21 5.53 0.2071 0.03103]) %Gpy= -.12528*tf([.31318 1],[.00001 1])*tf([5 1],[1 0],'iodelay',.25); %G=Gcy*Gpy W=LOGSPACE(log10(.1),log10(10),1000); ii=0; for kstarr=.1:.001:1; ii=ii+1; kstar(ii)=kstarr; mag_N(ii)=-20*log10(8*kstarr/(pi^2)); ph_N(ii)=(180/pi)*acos(kstarr) - 180; end figure(1) nichols(G,W) grid on hold on plot(ph_N,mag_N,'r') axis([-200 -70 -20 30]) grid on w1=input('what freq (rad/s) range? Low end =') w2=input('high end =') W1=LOGSPACE(log10(w1),log10(w2),1000); [mag_check1,phase_check1] = NICHOLS(G,W1); mag_check=squeeze(mag_check1); phase_check=squeeze(phase_check1); check = polyfit(phase_check,mag_check,9) swoosh = polyfit(ph_N,mag_N,9) jj=1 for phase=-100:-.1:-170; checkmagdB(jj)=20*log10(polyval(check,phase)); checkphase(jj)=phase; jj=jj+1; end jj=1; for phase=-100:-.1:-170; swooshmag(jj)=polyval(swoosh,phase); swooshphase(jj)=phase; jj=jj+1; end dB=swooshmag-checkmagdB; [Gap,index] = min(dB) inc=10^(Gap/20) nichols(G*inc,W)
A-6
hold on axis([-200 -70 -20 30]) omega=input('just touchin omega') [magtouch,phasetouch]=bode(G,omega) ii=1; for kstarr=.1:.001:1; if ph_N(ii)>phasetouch kstar=kstarr; end ii=ii+1; end kstar maxdeg=30 %Compute Gap Criterion for rate limits 15 deg/sec, 30 deg/sec and 60 deg/sec GapCriterion15=inc*(pi*15/(2*omega*kstar*maxdeg)) GapCriterion30=inc*(pi*30/(2*omega*kstar*maxdeg)) GapCriterion60=inc*(pi*60/(2*omega*kstar*maxdeg)) 'done'
B-1
Appendix B. State Space Matrices for Project MAX GAP
This appendix lists the state space matrices for the F-16 dynamics as well as the
bare aircraft dynamics of each case for Projects HAVE PREVENT, HAVE OLOP and
MAX GAP.
F-16 Dynamics a = b = c = d = Modified Matrices The following matrices are the pole placement results (A = a + b*kinner). HAVE PREVENT Case A A = Case B A = Case C A =
Data Basis: 15K ft PA, 300 KIASTest A/C: NF-16D - # 86-00048Acquisition System: Hand HeldConfiguration: Cruise / VSS EngagedTest Dates: 20-22 October 2003
Figure D-49. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case B, 15 deg/sec
2
4
3
2
5
0
1
2
3
4
5
6
1 2 2 2 3Evaluation Pilot
PIO
Ten
decy
Rat
ing
Data Basis: 15K ft PA, 300 KIASTest A/C: NF-16D - # 86-00048Acquisition System: Hand HeldConfiguration: Cruise / VSS EngagedTest Dates: 20-22 October 2003
Figure D-50. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case B, 30 deg/sec
D-27
2
3 3
2
4
0
1
2
3
4
5
6
1 1 2 2 3Evaluation Pilot
PIO
Ten
decy
Rat
ing
Data Basis: 15K ft PA, 300 KIASTest A/C: NF-16D - # 86-00048Acquisition System: Hand HeldConfiguration: Cruise / VSS EngagedTest Dates: 20-22 October 2003
Figure D-51. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case B, 30 deg/sec
4
3
4 4
0
1
2
3
4
5
6
1 1 1 3Evaluation Pilot
PIO
Ten
decy
Rat
ing
Data Basis: 15K ft PA, 300 KIASTest A/C: NF-16D - # 86-00048Acquisition System: Hand HeldConfiguration: Cruise / VSS EngagedTest Dates: 20-22 October 2003
Figure D-52. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case N, 15 deg/sec
D-28
2
3
0
1
2
3
4
5
6
3 3Evaluation Pilot
PIO
Ten
decy
Rat
ing
Data Basis: 15K ft PA, 300 KIASTest A/C: NF-16D - # 86-00048Acquisition System: Hand HeldConfiguration: Cruise / VSS EngagedTest Dates: 20-22 October 2003
Figure D-53. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case N, 30 deg/sec
3 3
4
2 2
0
1
2
3
4
5
6
1 1 1 2 2Evaluation Pilot
PIO
Ten
decy
Rat
ing
Data Basis: 15K ft PA, 300 KIASTest A/C: NF-16D - # 86-00048Acquisition System: Hand HeldConfiguration: Cruise / VSS EngagedTest Dates: 20-22 October 2003
Figure D-54. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case N, 60 deg/sec
D-29
4
5
4 4
5
0
1
2
3
4
5
6
1Evaluation Pilot
PIO
Ten
decy
Rat
ing
Data Basis: 15K ft PA, 300 KIASTest A/C: NF-16D - # 86-00048Acquisition System: Hand HeldConfiguration: Cruise / VSS EngagedTest Dates: 20-22 October 2003
Figure D-55. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case W, 15 deg/sec
2
5
2 2
0
1
2
3
4
5
6
2 2 2 3Evaluation Pilot
PIO
Ten
decy
Rat
ing
Data Basis: 15K ft PA, 300 KIASTest A/C: NF-16D - # 86-00048Acquisition System: Hand HeldConfiguration: Cruise / VSS EngagedTest Dates: 20-22 October 2003
Figure D-56. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case W, 30 deg/sec
D-30
2 2
0
1
2
3
4
5
6
3 3Evaluation Pilot
PIO
Ten
decy
Rat
ing
Data Basis: 15K ft PA, 300 KIASTest A/C: NF-16D - # 86-00048Acquisition System: Hand HeldConfiguration: Cruise / VSS EngagedTest Dates: 20-22 October 2003
Figure D-57. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case W, 60 deg/sec
2
4
5
4
0
1
2
3
4
5
6
1 1 2 2Evaluation Pilot
PIO
Ten
decy
Rat
ing
Data Basis: 15K ft PA, 300 KIASTest A/C: NF-16D - # 86-00048Acquisition System: Hand HeldConfiguration: Cruise / VSS EngagedTest Dates: 20-22 October 2003
Figure D-58. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case Y, 15 deg/sec
Om
itted
D-31
3
2
4
0
1
2
3
4
5
6
1 1 3Evaluation Pilot
PIO
Ten
decy
Rat
ing
Data Basis: 15K ft PA, 300 KIASTest A/C: NF-16D - # 86-00048Acquisition System: Hand HeldConfiguration: Cruise / VSS EngagedTest Dates: 20-22 October 2003
Figure D-59. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case Y, 30 deg/sec
2
3 3
2
3
0
1
2
3
4
5
6
1 1 1 2 3Evaluation Pilot
PIO
Ten
decy
Rat
ing
Data Basis: 15K ft PA, 300 KIASTest A/C: NF-16D - # 86-00048Acquisition System: Hand HeldConfiguration: Cruise / VSS EngagedTest Dates: 20-22 October 2003
Figure D-60. VISTA Phase 3 Discrete HUD Pitch Tracking Data, Case Y, 60 deg/sec
BIB-1
Bibliography
Ahlgren, Jan. Report on the Accident Involving the JAS 39-1 Gripen February 2, 1989. T1-1E-89. Government Accident Investigation Board (SHK), Sweden, March 1989. Anderson, Mark R. and Anthony B. Page. Multivariable Analysis of Pilot-in-the-Loop Oscillations. AIAA-95-3203-CP. Reston, VA: American Institute of Aeronautics and Astronautics, Inc., 1995. Brown, Frank and others. Flying Qualities Testing. Air Force Flight Test Center, Edwards AFB CA, 20 February 2002. Department of Defense. Flying Qualities of Piloted Aircraft. MIL-HDBK-1797. Washington: GPO, 19 December 1997. Dornheim, Michael A. “Boeing Corrects Several 777 PIOs,” Aviation Week and Space Technology, 142 (19): 32–33 (1995). Doman, David B. and Lori Ann Foringer. Interactive Flying Qualities Toolbox for Matlab, Version 2.0. Computer Software. USAF Wright Laboratory, Flight Dynamics Directorate, Flight Control Division, Flying Qualities Section, Wright- Patterson AFB OH, August 1996. Duda, Holger. Effects of Rate Limiting Elements in Flight Control Systems – A New PIO- Criterion. AIAA-95-3204-CP. Reston, VA: American Institute of Aeronautics and Astronautics, Inc., 1995. Gilbreath, Greg and others. A Limited Evaluation of a Pilot-Induced Oscillation Prediction Criterion (HAVE OLOP). AFFTC TIM-00-07. Air Force Flight Test Center, Edwards AFB, CA, December 2000. Gilbreath, Gregory P. Prediction of Pilot-Induced Oscillations (PIO) Due to Actuator Rate Limiting Using the Open-Loop Onset Point (OLOP) Criterio. MS Thesis, AFIT/GAE/ENY/01M-02, Graduate School of Engineering and Management, Air Force Institute of Technology (AU), Wright-Patterson AFB OH, March 2001. Hanley, James G. and others. Comparison of Nonlinear Algorithms in the Prevention of Pilot-Induced Oscillations Caused by Actuator Rate Limiting (Project HAVE PREVENT). Air Force Flight Test Center, Edwards AFB CA, December 2002 (ADA410173).
BIB-2
Hanley, James G. A Comparison of Nonlinear Algorithms to Prevent Pilot-Induced Oscillations Caused by Actuator Rate Limiting. MS thesis, AFIT/GAE/ENY/ 03-4. Graduate School of Engineering and Management, Air Force Institute of Technology (AU), Wright-Patterson AFB OH, March 2003 (ADA413992). Hodgkinson, John. Aircraft Handling Qualities. Reston, VA: American Institute of Aeronautics and Astronautics, Incorporated, 1999. Jensen, Clas and others. “JAS 39 Gripen EFCS: How to Deal with Rate Limiting.” Unpublished report. SAAB, Sweden, Undated. Klyde, David H. and others. Unified Pilot-Induced Oscillation Theory, Volume 1: PIO Analysis with Linear and Nonlinear Effective Vehicle Characteristics, Including Rate Limiting. WL-TR-96-3028. Air Force Research Laboratories, Wright- Patterson AFB OH, December 1995. Liebst, Brad S. Class Handout, MECH 629, Aircraft Handling Qualities and Performance. Graduate School of Engineering and Management, Air Force Institute of Technology, Wright-Patterson AFB OH, July 2001. Liebst, Brad S. and others. “Nonlinear Pre-filter to Prevent Pilot-Induced Oscillations Due to Actuator Rate Limiting.” AIAA Journal of Guidance, Control, and Dynamics, Vol. 25, No. 4: 740-747 (July 2002). MatlabTM /SimulinkTM. Versions 6.1 and 6.5. Computer software. The Mathworks, Inc., Natick MA, 2001. McRuer, Duane T. Pilot-Induced Oscillations and Human Dynamic Behavior. NASA Contractor Report 4683. Hawthorne CA: Systems Technology, Inc., 1995 Microsoft Excel. Version 10.3506.3501 SP-1. Computer Software. Microsoft Corporation, Redmond WA, 2001. Mitchell, David G. and Roger H. Hoh. Development of a Unified Method to Predict Tendencies for Pilot-Induced Oscillations. WL-TR-95-3049. Air Force Research Laboratories, Wright- Patterson AFB OH, June 1995. Mitchell, David G. and David H Klyde. A Critical Examination of PIO Prediction Criteria. AIAA-98-4335. Reston, VA: American Institute of Aeronautics and Astronautics, Inc., 1998. Preston, Jeff D. and others. Unified Pilot-Induced Oscillation Theory, Volume 2: Pilot- Induced Oscillation Criteria Applied to Several McDonnell Douglas Aircraft. WL-TR-96-3029. Air Force Research Laboratories, Wright-Patterson AFB OH, December 1995 (ADB212157).
BIB-3
Slotine, Jean-Jacques E. and Weiping Li. Applied Nonlinear Control. Upper Saddle River, New Jersey: Prentice Hall, 1991. Wheeler, Anthony J. and Ahmand R. Ganji. Introduction to Engineering Experimentation. Upper Saddle River, New Jersey: Prentice Hall, 1996. Witte, Joel B. and others. An Investigation Relating Longitudinal Pilot-induced Oscillation Tendency Ratings to Describing Function Predictions for Rate-limited Actuators (Project MAX GAP). Air Force Flight Test Center, Edwards AFB CA, December 2003.
VITA-1
Vita
Major Joel B. Witte was born in Ft Worth, Texas. He graduated from Burleson
High School, Burleson, Texas in 1986. He earned a Bachelor of Science Degree in
Aerospace Engineering from Texas A&M University in 1990.
After graduation, Major Witte entered Undergraduate Pilot Training at Reese
AFB, Texas in January 1992. He received his wings one year later and was assigned to
fly the C-27A at Howard AFB, Panama. His follow-on assignments included flying the
C-141B and C-17A at Charleston AFB, South Carolina and McChord AFB, Washington.
He has accumulated more than 2800 hours flying time.
Major Witte was selected for the joint Air Force Institute of Technology/USAF
Test Pilot School program in February 2001 and began classes at AFIT in September.
After graduation from AFIT/TPS, he will test Special Operations C-130s at Hurlburt
Field, Florida.
Major Witte is a graduate of Test Pilot School, Class 03A – “The Centurions.”
VITA-1
REPORT DOCUMENTATION PAGE Form Approved OMB No. 074-0188
The public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of the collection of information, including suggestions for reducing this burden to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports (0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to an penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number. PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS. 1. REPORT DATE (DD-MM-YYYY)
03-2004 2. REPORT TYPE
Master’s Thesis 3. DATES COVERED (From – To)
Apr 2002 – Mar 2004 5a. CONTRACT NUMBER
5b. GRANT NUMBER
4. TITLE AND SUBTITLE 5.
AN INVESTIGATION RELATING LONGITUDINAL PILOT-INDUCED OSCILLATION TENDENCY RATING TO DESCRIBING FUNCTION PREDICTIONS FOR RATE-LIMITED
ACTUATORS
5c. PROGRAM ELEMENT NUMBER
5d. PROJECT NUMBER 5e. TASK NUMBER
6. AUTHOR(S) Witte, Joel B. Major, USAF
5f. WORK UNIT NUMBER
7. PERFORMING ORGANIZATION NAMES(S) AND ADDRESS(S) Air Force Institute of Technology Graduate School of Engineering and Management (AFIT/EN) 2950 P Street, Building 640 WPAFB OH 45433-7765
8. PERFORMING ORGANIZATION REPORT NUMBER AFIT/GAE/ENY/04-M16
10. SPONSOR/MONITOR’S ACRONYM(S)
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) USAF TPS 220 Wolfe Ave Edwards AFB CA 93524 DSN: 527-3000 11. SPONSOR/MONITOR’S
REPORT NUMBER(S)
12. DISTRIBUTION/AVAILABILITY STATEMENT APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED
13. SUPPLEMENTARY NOTES 14. ABSTRACT The purpose of this study was to investigate pilot-induced oscillations (PIO) and determine a method by which PIO tendency rating could be predicted. In particular, longitudinal PIO in the presence of rate-limited actuators were singled out for examination. Sinusoidal input/triangular output describing function techniques using Nichols charts were used. A new criterion dubbed Gap Criterion was calculated for PIO sensitivity. This criterion consists of the product of pilot gain necessary to cause PIO and the normalized maximum amplitude of the commanded actuator. These results were paired with simulator and flight test PIO tendency rating data. The PIO rating scale used was the PIO tendency classification of MIL-HDBK-1797. This concept was applied to two historical test databases, HAVE PREVENT and HAVE OLOP. Additional PIO data was gathered in the Large Amplitude Multimode Aerospace Simulator (LAMARS) at the Air Force Research Laboratory (AFRL), Wright-Patterson AFB, Ohio and the USAF NF-16D Variable In-flight Stability Test Aircraft (VISTA) at Edwards AFB, California. Correlation between PIO tendency rating and Gap Criterion was determined for each dataset. Most datasets exceeded a confidence level of 95% that a correlation existed. Follow-on analysis for best curve fit was also accomplished with a logarithmic fit deemed best. Datasets were combined with success to demonstrate the universality of the Gap Criterion for correlating PIO tendency ratings. 15. SUBJECT TERMS handling qualities pilot-induced oscillations pilot-induced oscillation prediction rate limiting flight controls
16. SECURITY CLASSIFICATION OF:
19a. NAME OF RESPONSIBLE PERSON Dr Brad Liebst, AFIT ENY
REPORT U
ABSTRACT U
c. THIS PAGE U
17. LIMITATION OF ABSTRACT UU
18. NUMBER OF PAGES 154 19b. TELEPHONE NUMBER (Include area code)
(937) 255-3636, ext 4636
VITA-1
Standard Form 298 (Rev: 8-98) Prescribed by ANSI Std Z39-18